Decoding FIP loss equation

[HariniNarasimhan]

Hi @Guruprasad93 ,
Thanks for this solution to solve catastrophic forgetting and it is very interesting to learn. Please help me to decode the below which I find it difficult to understand.

loss1 = torch.sum(-torch.log(torch.sum(torch.sqrt(output*output_ori+epsAdd), axis=1)))

FlexibleMachineLearning/VIT/train_fip.py

Line 223 in 8374c5a

loss1 = torch.sum(-torch.log(torch.sum(torch.sqrt(output*output_ori+epsAdd), axis=1)))

The loss1 is computed but not used for the loss to back propagate. Can you please help me with reasoning?